1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This class implements the parser for assembly files.
12 //===----------------------------------------------------------------------===//
14 #include "llvm/ADT/APFloat.h"
15 #include "llvm/ADT/SmallString.h"
16 #include "llvm/ADT/StringMap.h"
17 #include "llvm/ADT/StringSwitch.h"
18 #include "llvm/ADT/Twine.h"
19 #include "llvm/MC/MCAsmInfo.h"
20 #include "llvm/MC/MCContext.h"
21 #include "llvm/MC/MCExpr.h"
22 #include "llvm/MC/MCParser/AsmCond.h"
23 #include "llvm/MC/MCParser/AsmLexer.h"
24 #include "llvm/MC/MCParser/MCAsmParser.h"
25 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
26 #include "llvm/MC/MCSectionMachO.h"
27 #include "llvm/MC/MCStreamer.h"
28 #include "llvm/MC/MCSymbol.h"
29 #include "llvm/MC/MCDwarf.h"
30 #include "llvm/Support/MemoryBuffer.h"
31 #include "llvm/Support/SourceMgr.h"
32 #include "llvm/Support/raw_ostream.h"
33 #include "llvm/Target/TargetAsmInfo.h"
34 #include "llvm/Target/TargetAsmParser.h"
41 /// \brief Helper class for tracking macro definitions.
47 Macro(StringRef N, StringRef B) : Name(N), Body(B) {}
50 /// \brief Helper class for storing information about an active macro
52 struct MacroInstantiation {
53 /// The macro being instantiated.
54 const Macro *TheMacro;
56 /// The macro instantiation with substitutions.
57 MemoryBuffer *Instantiation;
59 /// The location of the instantiation.
60 SMLoc InstantiationLoc;
62 /// The location where parsing should resume upon instantiation completion.
66 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
67 const std::vector<std::vector<AsmToken> > &A);
70 /// \brief The concrete assembly parser instance.
71 class AsmParser : public MCAsmParser {
72 friend class GenericAsmParser;
74 AsmParser(const AsmParser &); // DO NOT IMPLEMENT
75 void operator=(const AsmParser &); // DO NOT IMPLEMENT
81 MCAsmParserExtension *GenericParser;
82 MCAsmParserExtension *PlatformParser;
84 /// This is the current buffer index we're lexing from as managed by the
89 std::vector<AsmCond> TheCondStack;
91 /// DirectiveMap - This is a table handlers for directives. Each handler is
92 /// invoked after the directive identifier is read and is responsible for
93 /// parsing and validating the rest of the directive. The handler is passed
94 /// in the directive name and the location of the directive keyword.
95 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap;
97 /// MacroMap - Map of currently defined macros.
98 StringMap<Macro*> MacroMap;
100 /// ActiveMacros - Stack of active macro instantiations.
101 std::vector<MacroInstantiation*> ActiveMacros;
103 /// Boolean tracking whether macro substitution is enabled.
104 unsigned MacrosEnabled : 1;
106 /// Flag tracking whether any errors have been encountered.
107 unsigned HadError : 1;
110 AsmParser(const Target &T, SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
111 const MCAsmInfo &MAI);
114 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false);
116 void AddDirectiveHandler(MCAsmParserExtension *Object,
118 DirectiveHandler Handler) {
119 DirectiveMap[Directive] = std::make_pair(Object, Handler);
123 /// @name MCAsmParser Interface
126 virtual SourceMgr &getSourceManager() { return SrcMgr; }
127 virtual MCAsmLexer &getLexer() { return Lexer; }
128 virtual MCContext &getContext() { return Ctx; }
129 virtual MCStreamer &getStreamer() { return Out; }
131 virtual void Warning(SMLoc L, const Twine &Meg);
132 virtual bool Error(SMLoc L, const Twine &Msg);
134 const AsmToken &Lex();
136 bool ParseExpression(const MCExpr *&Res);
137 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc);
138 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc);
139 virtual bool ParseAbsoluteExpression(int64_t &Res);
144 void CheckForValidSection();
146 bool ParseStatement();
148 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M);
149 void HandleMacroExit();
151 void PrintMacroInstantiations();
152 void PrintMessage(SMLoc Loc, const Twine &Msg, const char *Type) const {
153 SrcMgr.PrintMessage(Loc, Msg, Type);
156 /// EnterIncludeFile - Enter the specified file. This returns true on failure.
157 bool EnterIncludeFile(const std::string &Filename);
159 /// \brief Reset the current lexer position to that given by \arg Loc. The
160 /// current token is not set; clients should ensure Lex() is called
162 void JumpToLoc(SMLoc Loc);
164 void EatToEndOfStatement();
166 /// \brief Parse up to the end of statement and a return the contents from the
167 /// current token until the end of the statement; the current token on exit
168 /// will be either the EndOfStatement or EOF.
169 StringRef ParseStringToEndOfStatement();
171 bool ParseAssignment(StringRef Name, bool allow_redef);
173 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc);
174 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
175 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
176 bool ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
178 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier)
179 /// and set \arg Res to the identifier contents.
180 bool ParseIdentifier(StringRef &Res);
182 // Directive Parsing.
184 // ".ascii", ".asciiz", ".string"
185 bool ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
186 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ...
187 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ...
188 bool ParseDirectiveFill(); // ".fill"
189 bool ParseDirectiveSpace(); // ".space"
190 bool ParseDirectiveZero(); // ".zero"
191 bool ParseDirectiveSet(StringRef IDVal, bool allow_redef); // ".set", ".equ", ".equiv"
192 bool ParseDirectiveOrg(); // ".org"
193 // ".align{,32}", ".p2align{,w,l}"
194 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize);
196 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which
197 /// accepts a single symbol (which should be a label or an external).
198 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr);
200 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
202 bool ParseDirectiveAbort(); // ".abort"
203 bool ParseDirectiveInclude(); // ".include"
205 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if"
206 // ".ifdef" or ".ifndef", depending on expect_defined
207 bool ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
208 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
209 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else"
210 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
212 /// ParseEscapedString - Parse the current token as a string which may include
213 /// escaped characters and return the string contents.
214 bool ParseEscapedString(std::string &Data);
216 const MCExpr *ApplyModifierToExpr(const MCExpr *E,
217 MCSymbolRefExpr::VariantKind Variant);
220 /// \brief Generic implementations of directive handling, etc. which is shared
221 /// (or the default, at least) for all assembler parser.
222 class GenericAsmParser : public MCAsmParserExtension {
223 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)>
224 void AddDirectiveHandler(StringRef Directive) {
225 getParser().AddDirectiveHandler(this, Directive,
226 HandleDirective<GenericAsmParser, Handler>);
229 GenericAsmParser() {}
231 AsmParser &getParser() {
232 return (AsmParser&) this->MCAsmParserExtension::getParser();
235 virtual void Initialize(MCAsmParser &Parser) {
236 // Call the base implementation.
237 this->MCAsmParserExtension::Initialize(Parser);
239 // Debugging directives.
240 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file");
241 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line");
242 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc");
243 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs");
246 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIStartProc>(
248 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIEndProc>(
250 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfa>(
252 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaOffset>(
253 ".cfi_def_cfa_offset");
254 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaRegister>(
255 ".cfi_def_cfa_register");
256 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIOffset>(
259 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_personality");
261 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_lsda");
263 &GenericAsmParser::ParseDirectiveCFIRememberState>(".cfi_remember_state");
265 &GenericAsmParser::ParseDirectiveCFIRestoreState>(".cfi_restore_state");
268 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
270 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
272 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
273 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
274 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
276 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
277 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
280 bool ParseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
282 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc);
283 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc);
284 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc);
285 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc);
286 bool ParseDirectiveCFIStartProc(StringRef, SMLoc DirectiveLoc);
287 bool ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc);
288 bool ParseDirectiveCFIDefCfa(StringRef, SMLoc DirectiveLoc);
289 bool ParseDirectiveCFIDefCfaOffset(StringRef, SMLoc DirectiveLoc);
290 bool ParseDirectiveCFIDefCfaRegister(StringRef, SMLoc DirectiveLoc);
291 bool ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc);
292 bool ParseDirectiveCFIPersonalityOrLsda(StringRef, SMLoc DirectiveLoc);
293 bool ParseDirectiveCFIRememberState(StringRef, SMLoc DirectiveLoc);
294 bool ParseDirectiveCFIRestoreState(StringRef, SMLoc DirectiveLoc);
296 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
297 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
298 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
300 bool ParseDirectiveLEB128(StringRef, SMLoc);
307 extern MCAsmParserExtension *createDarwinAsmParser();
308 extern MCAsmParserExtension *createELFAsmParser();
309 extern MCAsmParserExtension *createCOFFAsmParser();
313 enum { DEFAULT_ADDRSPACE = 0 };
315 AsmParser::AsmParser(const Target &T, SourceMgr &_SM, MCContext &_Ctx,
316 MCStreamer &_Out, const MCAsmInfo &_MAI)
317 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), SrcMgr(_SM),
318 GenericParser(new GenericAsmParser), PlatformParser(0),
319 CurBuffer(0), MacrosEnabled(true) {
320 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
322 // Initialize the generic parser.
323 GenericParser->Initialize(*this);
325 // Initialize the platform / file format parser.
327 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
329 if (_MAI.hasMicrosoftFastStdCallMangling()) {
330 PlatformParser = createCOFFAsmParser();
331 PlatformParser->Initialize(*this);
332 } else if (_MAI.hasSubsectionsViaSymbols()) {
333 PlatformParser = createDarwinAsmParser();
334 PlatformParser->Initialize(*this);
336 PlatformParser = createELFAsmParser();
337 PlatformParser->Initialize(*this);
341 AsmParser::~AsmParser() {
342 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
344 // Destroy any macros.
345 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
346 ie = MacroMap.end(); it != ie; ++it)
347 delete it->getValue();
349 delete PlatformParser;
350 delete GenericParser;
353 void AsmParser::PrintMacroInstantiations() {
354 // Print the active macro instantiation stack.
355 for (std::vector<MacroInstantiation*>::const_reverse_iterator
356 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
357 PrintMessage((*it)->InstantiationLoc, "while in macro instantiation",
361 void AsmParser::Warning(SMLoc L, const Twine &Msg) {
362 PrintMessage(L, Msg, "warning");
363 PrintMacroInstantiations();
366 bool AsmParser::Error(SMLoc L, const Twine &Msg) {
368 PrintMessage(L, Msg, "error");
369 PrintMacroInstantiations();
373 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
374 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc());
380 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
385 void AsmParser::JumpToLoc(SMLoc Loc) {
386 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
387 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
390 const AsmToken &AsmParser::Lex() {
391 const AsmToken *tok = &Lexer.Lex();
393 if (tok->is(AsmToken::Eof)) {
394 // If this is the end of an included file, pop the parent file off the
396 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
397 if (ParentIncludeLoc != SMLoc()) {
398 JumpToLoc(ParentIncludeLoc);
403 if (tok->is(AsmToken::Error))
404 Error(Lexer.getErrLoc(), Lexer.getErr());
409 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
410 // Create the initial section, if requested.
411 if (!NoInitialTextSection)
418 AsmCond StartingCondState = TheCondState;
420 // While we have input, parse each statement.
421 while (Lexer.isNot(AsmToken::Eof)) {
422 if (!ParseStatement()) continue;
424 // We had an error, validate that one was emitted and recover by skipping to
426 assert(HadError && "Parse statement returned an error, but none emitted!");
427 EatToEndOfStatement();
430 if (TheCondState.TheCond != StartingCondState.TheCond ||
431 TheCondState.Ignore != StartingCondState.Ignore)
432 return TokError("unmatched .ifs or .elses");
434 // Check to see there are no empty DwarfFile slots.
435 const std::vector<MCDwarfFile *> &MCDwarfFiles =
436 getContext().getMCDwarfFiles();
437 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
438 if (!MCDwarfFiles[i])
439 TokError("unassigned file number: " + Twine(i) + " for .file directives");
442 // Finalize the output stream if there are no errors and if the client wants
444 if (!HadError && !NoFinalize)
450 void AsmParser::CheckForValidSection() {
451 if (!getStreamer().getCurrentSection()) {
452 TokError("expected section directive before assembly directive");
453 Out.SwitchSection(Ctx.getMachOSection(
455 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
456 0, SectionKind::getText()));
460 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
461 void AsmParser::EatToEndOfStatement() {
462 while (Lexer.isNot(AsmToken::EndOfStatement) &&
463 Lexer.isNot(AsmToken::Eof))
467 if (Lexer.is(AsmToken::EndOfStatement))
471 StringRef AsmParser::ParseStringToEndOfStatement() {
472 const char *Start = getTok().getLoc().getPointer();
474 while (Lexer.isNot(AsmToken::EndOfStatement) &&
475 Lexer.isNot(AsmToken::Eof))
478 const char *End = getTok().getLoc().getPointer();
479 return StringRef(Start, End - Start);
482 /// ParseParenExpr - Parse a paren expression and return it.
483 /// NOTE: This assumes the leading '(' has already been consumed.
485 /// parenexpr ::= expr)
487 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
488 if (ParseExpression(Res)) return true;
489 if (Lexer.isNot(AsmToken::RParen))
490 return TokError("expected ')' in parentheses expression");
491 EndLoc = Lexer.getLoc();
496 /// ParseBracketExpr - Parse a bracket expression and return it.
497 /// NOTE: This assumes the leading '[' has already been consumed.
499 /// bracketexpr ::= expr]
501 bool AsmParser::ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
502 if (ParseExpression(Res)) return true;
503 if (Lexer.isNot(AsmToken::RBrac))
504 return TokError("expected ']' in brackets expression");
505 EndLoc = Lexer.getLoc();
510 /// ParsePrimaryExpr - Parse a primary expression and return it.
511 /// primaryexpr ::= (parenexpr
512 /// primaryexpr ::= symbol
513 /// primaryexpr ::= number
514 /// primaryexpr ::= '.'
515 /// primaryexpr ::= ~,+,- primaryexpr
516 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
517 switch (Lexer.getKind()) {
519 return TokError("unknown token in expression");
520 case AsmToken::Exclaim:
521 Lex(); // Eat the operator.
522 if (ParsePrimaryExpr(Res, EndLoc))
524 Res = MCUnaryExpr::CreateLNot(Res, getContext());
526 case AsmToken::Dollar:
527 case AsmToken::String:
528 case AsmToken::Identifier: {
529 EndLoc = Lexer.getLoc();
531 StringRef Identifier;
532 if (ParseIdentifier(Identifier))
535 // This is a symbol reference.
536 std::pair<StringRef, StringRef> Split = Identifier.split('@');
537 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
539 // Lookup the symbol variant if used.
540 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
541 if (Split.first.size() != Identifier.size()) {
542 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
543 if (Variant == MCSymbolRefExpr::VK_Invalid) {
544 Variant = MCSymbolRefExpr::VK_None;
545 return TokError("invalid variant '" + Split.second + "'");
549 // If this is an absolute variable reference, substitute it now to preserve
550 // semantics in the face of reassignment.
551 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
553 return Error(EndLoc, "unexpected modifier on variable reference");
555 Res = Sym->getVariableValue();
559 // Otherwise create a symbol ref.
560 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
563 case AsmToken::Integer: {
564 SMLoc Loc = getTok().getLoc();
565 int64_t IntVal = getTok().getIntVal();
566 Res = MCConstantExpr::Create(IntVal, getContext());
567 EndLoc = Lexer.getLoc();
569 // Look for 'b' or 'f' following an Integer as a directional label
570 if (Lexer.getKind() == AsmToken::Identifier) {
571 StringRef IDVal = getTok().getString();
572 if (IDVal == "f" || IDVal == "b"){
573 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
574 IDVal == "f" ? 1 : 0);
575 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
577 if(IDVal == "b" && Sym->isUndefined())
578 return Error(Loc, "invalid reference to undefined symbol");
579 EndLoc = Lexer.getLoc();
580 Lex(); // Eat identifier.
585 case AsmToken::Real: {
586 APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
587 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
588 Res = MCConstantExpr::Create(IntVal, getContext());
592 case AsmToken::Dot: {
593 // This is a '.' reference, which references the current PC. Emit a
594 // temporary label to the streamer and refer to it.
595 MCSymbol *Sym = Ctx.CreateTempSymbol();
597 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
598 EndLoc = Lexer.getLoc();
599 Lex(); // Eat identifier.
602 case AsmToken::LParen:
603 Lex(); // Eat the '('.
604 return ParseParenExpr(Res, EndLoc);
605 case AsmToken::LBrac:
606 if (!PlatformParser->HasBracketExpressions())
607 return TokError("brackets expression not supported on this target");
608 Lex(); // Eat the '['.
609 return ParseBracketExpr(Res, EndLoc);
610 case AsmToken::Minus:
611 Lex(); // Eat the operator.
612 if (ParsePrimaryExpr(Res, EndLoc))
614 Res = MCUnaryExpr::CreateMinus(Res, getContext());
617 Lex(); // Eat the operator.
618 if (ParsePrimaryExpr(Res, EndLoc))
620 Res = MCUnaryExpr::CreatePlus(Res, getContext());
622 case AsmToken::Tilde:
623 Lex(); // Eat the operator.
624 if (ParsePrimaryExpr(Res, EndLoc))
626 Res = MCUnaryExpr::CreateNot(Res, getContext());
631 bool AsmParser::ParseExpression(const MCExpr *&Res) {
633 return ParseExpression(Res, EndLoc);
637 AsmParser::ApplyModifierToExpr(const MCExpr *E,
638 MCSymbolRefExpr::VariantKind Variant) {
639 // Recurse over the given expression, rebuilding it to apply the given variant
640 // if there is exactly one symbol.
641 switch (E->getKind()) {
643 case MCExpr::Constant:
646 case MCExpr::SymbolRef: {
647 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
649 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
650 TokError("invalid variant on expression '" +
651 getTok().getIdentifier() + "' (already modified)");
655 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
658 case MCExpr::Unary: {
659 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
660 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
663 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
666 case MCExpr::Binary: {
667 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
668 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
669 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
674 if (!LHS) LHS = BE->getLHS();
675 if (!RHS) RHS = BE->getRHS();
677 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
681 assert(0 && "Invalid expression kind!");
685 /// ParseExpression - Parse an expression and return it.
687 /// expr ::= expr +,- expr -> lowest.
688 /// expr ::= expr |,^,&,! expr -> middle.
689 /// expr ::= expr *,/,%,<<,>> expr -> highest.
690 /// expr ::= primaryexpr
692 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
693 // Parse the expression.
695 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
698 // As a special case, we support 'a op b @ modifier' by rewriting the
699 // expression to include the modifier. This is inefficient, but in general we
700 // expect users to use 'a@modifier op b'.
701 if (Lexer.getKind() == AsmToken::At) {
704 if (Lexer.isNot(AsmToken::Identifier))
705 return TokError("unexpected symbol modifier following '@'");
707 MCSymbolRefExpr::VariantKind Variant =
708 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
709 if (Variant == MCSymbolRefExpr::VK_Invalid)
710 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
712 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
714 return TokError("invalid modifier '" + getTok().getIdentifier() +
715 "' (no symbols present)");
723 // Try to constant fold it up front, if possible.
725 if (Res->EvaluateAsAbsolute(Value))
726 Res = MCConstantExpr::Create(Value, getContext());
731 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
733 return ParseParenExpr(Res, EndLoc) ||
734 ParseBinOpRHS(1, Res, EndLoc);
737 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
740 SMLoc StartLoc = Lexer.getLoc();
741 if (ParseExpression(Expr))
744 if (!Expr->EvaluateAsAbsolute(Res))
745 return Error(StartLoc, "expected absolute expression");
750 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
751 MCBinaryExpr::Opcode &Kind) {
754 return 0; // not a binop.
756 // Lowest Precedence: &&, ||, @
757 case AsmToken::AmpAmp:
758 Kind = MCBinaryExpr::LAnd;
760 case AsmToken::PipePipe:
761 Kind = MCBinaryExpr::LOr;
765 // Low Precedence: |, &, ^
767 // FIXME: gas seems to support '!' as an infix operator?
769 Kind = MCBinaryExpr::Or;
771 case AsmToken::Caret:
772 Kind = MCBinaryExpr::Xor;
775 Kind = MCBinaryExpr::And;
778 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
779 case AsmToken::EqualEqual:
780 Kind = MCBinaryExpr::EQ;
782 case AsmToken::ExclaimEqual:
783 case AsmToken::LessGreater:
784 Kind = MCBinaryExpr::NE;
787 Kind = MCBinaryExpr::LT;
789 case AsmToken::LessEqual:
790 Kind = MCBinaryExpr::LTE;
792 case AsmToken::Greater:
793 Kind = MCBinaryExpr::GT;
795 case AsmToken::GreaterEqual:
796 Kind = MCBinaryExpr::GTE;
799 // High Intermediate Precedence: +, -
801 Kind = MCBinaryExpr::Add;
803 case AsmToken::Minus:
804 Kind = MCBinaryExpr::Sub;
807 // Highest Precedence: *, /, %, <<, >>
809 Kind = MCBinaryExpr::Mul;
811 case AsmToken::Slash:
812 Kind = MCBinaryExpr::Div;
814 case AsmToken::Percent:
815 Kind = MCBinaryExpr::Mod;
817 case AsmToken::LessLess:
818 Kind = MCBinaryExpr::Shl;
820 case AsmToken::GreaterGreater:
821 Kind = MCBinaryExpr::Shr;
827 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
828 /// Res contains the LHS of the expression on input.
829 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
832 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
833 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
835 // If the next token is lower precedence than we are allowed to eat, return
836 // successfully with what we ate already.
837 if (TokPrec < Precedence)
842 // Eat the next primary expression.
844 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
846 // If BinOp binds less tightly with RHS than the operator after RHS, let
847 // the pending operator take RHS as its LHS.
848 MCBinaryExpr::Opcode Dummy;
849 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
850 if (TokPrec < NextTokPrec) {
851 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true;
854 // Merge LHS and RHS according to operator.
855 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
863 /// ::= EndOfStatement
864 /// ::= Label* Directive ...Operands... EndOfStatement
865 /// ::= Label* Identifier OperandList* EndOfStatement
866 bool AsmParser::ParseStatement() {
867 if (Lexer.is(AsmToken::EndOfStatement)) {
873 // Statements always start with an identifier or are a full line comment.
874 AsmToken ID = getTok();
875 SMLoc IDLoc = ID.getLoc();
877 int64_t LocalLabelVal = -1;
878 // A full line comment is a '#' as the first token.
879 if (Lexer.is(AsmToken::Hash)) {
880 EatToEndOfStatement();
883 // Allow an integer followed by a ':' as a directional local label.
884 if (Lexer.is(AsmToken::Integer)) {
885 LocalLabelVal = getTok().getIntVal();
886 if (LocalLabelVal < 0) {
887 if (!TheCondState.Ignore)
888 return TokError("unexpected token at start of statement");
892 IDVal = getTok().getString();
893 Lex(); // Consume the integer token to be used as an identifier token.
894 if (Lexer.getKind() != AsmToken::Colon) {
895 if (!TheCondState.Ignore)
896 return TokError("unexpected token at start of statement");
900 else if (ParseIdentifier(IDVal)) {
901 if (!TheCondState.Ignore)
902 return TokError("unexpected token at start of statement");
906 // Handle conditional assembly here before checking for skipping. We
907 // have to do this so that .endif isn't skipped in a ".if 0" block for
910 return ParseDirectiveIf(IDLoc);
911 if (IDVal == ".ifdef")
912 return ParseDirectiveIfdef(IDLoc, true);
913 if (IDVal == ".ifndef" || IDVal == ".ifnotdef")
914 return ParseDirectiveIfdef(IDLoc, false);
915 if (IDVal == ".elseif")
916 return ParseDirectiveElseIf(IDLoc);
917 if (IDVal == ".else")
918 return ParseDirectiveElse(IDLoc);
919 if (IDVal == ".endif")
920 return ParseDirectiveEndIf(IDLoc);
922 // If we are in a ".if 0" block, ignore this statement.
923 if (TheCondState.Ignore) {
924 EatToEndOfStatement();
928 // FIXME: Recurse on local labels?
930 // See what kind of statement we have.
931 switch (Lexer.getKind()) {
932 case AsmToken::Colon: {
933 CheckForValidSection();
935 // identifier ':' -> Label.
938 // Diagnose attempt to use a variable as a label.
940 // FIXME: Diagnostics. Note the location of the definition as a label.
941 // FIXME: This doesn't diagnose assignment to a symbol which has been
942 // implicitly marked as external.
944 if (LocalLabelVal == -1)
945 Sym = getContext().GetOrCreateSymbol(IDVal);
947 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
948 if (!Sym->isUndefined() || Sym->isVariable())
949 return Error(IDLoc, "invalid symbol redefinition");
954 // Consume any end of statement token, if present, to avoid spurious
955 // AddBlankLine calls().
956 if (Lexer.is(AsmToken::EndOfStatement)) {
958 if (Lexer.is(AsmToken::Eof))
962 return ParseStatement();
965 case AsmToken::Equal:
966 // identifier '=' ... -> assignment statement
969 return ParseAssignment(IDVal, true);
971 default: // Normal instruction or directive.
975 // If macros are enabled, check to see if this is a macro instantiation.
977 if (const Macro *M = MacroMap.lookup(IDVal))
978 return HandleMacroEntry(IDVal, IDLoc, M);
980 // Otherwise, we have a normal instruction or directive.
981 if (IDVal[0] == '.') {
982 // Assembler features
983 if (IDVal == ".set" || IDVal == ".equ")
984 return ParseDirectiveSet(IDVal, true);
985 if (IDVal == ".equiv")
986 return ParseDirectiveSet(IDVal, false);
990 if (IDVal == ".ascii")
991 return ParseDirectiveAscii(IDVal, false);
992 if (IDVal == ".asciz" || IDVal == ".string")
993 return ParseDirectiveAscii(IDVal, true);
995 if (IDVal == ".byte")
996 return ParseDirectiveValue(1);
997 if (IDVal == ".short")
998 return ParseDirectiveValue(2);
999 if (IDVal == ".value")
1000 return ParseDirectiveValue(2);
1001 if (IDVal == ".2byte")
1002 return ParseDirectiveValue(2);
1003 if (IDVal == ".long")
1004 return ParseDirectiveValue(4);
1005 if (IDVal == ".int")
1006 return ParseDirectiveValue(4);
1007 if (IDVal == ".4byte")
1008 return ParseDirectiveValue(4);
1009 if (IDVal == ".quad")
1010 return ParseDirectiveValue(8);
1011 if (IDVal == ".8byte")
1012 return ParseDirectiveValue(8);
1013 if (IDVal == ".single" || IDVal == ".float")
1014 return ParseDirectiveRealValue(APFloat::IEEEsingle);
1015 if (IDVal == ".double")
1016 return ParseDirectiveRealValue(APFloat::IEEEdouble);
1018 if (IDVal == ".align") {
1019 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1020 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1022 if (IDVal == ".align32") {
1023 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1024 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
1026 if (IDVal == ".balign")
1027 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
1028 if (IDVal == ".balignw")
1029 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
1030 if (IDVal == ".balignl")
1031 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
1032 if (IDVal == ".p2align")
1033 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
1034 if (IDVal == ".p2alignw")
1035 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
1036 if (IDVal == ".p2alignl")
1037 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
1039 if (IDVal == ".org")
1040 return ParseDirectiveOrg();
1042 if (IDVal == ".fill")
1043 return ParseDirectiveFill();
1044 if (IDVal == ".space")
1045 return ParseDirectiveSpace();
1046 if (IDVal == ".zero")
1047 return ParseDirectiveZero();
1049 // Symbol attribute directives
1051 if (IDVal == ".globl" || IDVal == ".global")
1052 return ParseDirectiveSymbolAttribute(MCSA_Global);
1053 // ELF only? Should it be here?
1054 if (IDVal == ".local")
1055 return ParseDirectiveSymbolAttribute(MCSA_Local);
1056 if (IDVal == ".hidden")
1057 return ParseDirectiveSymbolAttribute(MCSA_Hidden);
1058 if (IDVal == ".indirect_symbol")
1059 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
1060 if (IDVal == ".internal")
1061 return ParseDirectiveSymbolAttribute(MCSA_Internal);
1062 if (IDVal == ".lazy_reference")
1063 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
1064 if (IDVal == ".no_dead_strip")
1065 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
1066 if (IDVal == ".symbol_resolver")
1067 return ParseDirectiveSymbolAttribute(MCSA_SymbolResolver);
1068 if (IDVal == ".private_extern")
1069 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
1070 if (IDVal == ".protected")
1071 return ParseDirectiveSymbolAttribute(MCSA_Protected);
1072 if (IDVal == ".reference")
1073 return ParseDirectiveSymbolAttribute(MCSA_Reference);
1074 if (IDVal == ".weak")
1075 return ParseDirectiveSymbolAttribute(MCSA_Weak);
1076 if (IDVal == ".weak_definition")
1077 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
1078 if (IDVal == ".weak_reference")
1079 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
1080 if (IDVal == ".weak_def_can_be_hidden")
1081 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1083 if (IDVal == ".comm")
1084 return ParseDirectiveComm(/*IsLocal=*/false);
1085 if (IDVal == ".lcomm")
1086 return ParseDirectiveComm(/*IsLocal=*/true);
1088 if (IDVal == ".abort")
1089 return ParseDirectiveAbort();
1090 if (IDVal == ".include")
1091 return ParseDirectiveInclude();
1093 if (IDVal == ".code16" || IDVal == ".code32" || IDVal == ".code64")
1094 return TokError(Twine(IDVal) + " not supported yet");
1096 // Look up the handler in the handler table.
1097 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1098 DirectiveMap.lookup(IDVal);
1100 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1102 // Target hook for parsing target specific directives.
1103 if (!getTargetParser().ParseDirective(ID))
1106 Warning(IDLoc, "ignoring directive for now");
1107 EatToEndOfStatement();
1111 CheckForValidSection();
1113 // Canonicalize the opcode to lower case.
1114 SmallString<128> Opcode;
1115 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1116 Opcode.push_back(tolower(IDVal[i]));
1118 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1119 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
1122 // Dump the parsed representation, if requested.
1123 if (getShowParsedOperands()) {
1124 SmallString<256> Str;
1125 raw_svector_ostream OS(Str);
1126 OS << "parsed instruction: [";
1127 for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
1130 ParsedOperands[i]->dump(OS);
1134 PrintMessage(IDLoc, OS.str(), "note");
1137 // If parsing succeeded, match the instruction.
1139 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands,
1142 // Free any parsed operands.
1143 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1144 delete ParsedOperands[i];
1146 // Don't skip the rest of the line, the instruction parser is responsible for
1151 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1152 const std::vector<std::vector<AsmToken> > &A)
1153 : TheMacro(M), InstantiationLoc(IL), ExitLoc(EL)
1155 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1156 // to hold the macro body with substitutions.
1157 SmallString<256> Buf;
1158 raw_svector_ostream OS(Buf);
1160 StringRef Body = M->Body;
1161 while (!Body.empty()) {
1162 // Scan for the next substitution.
1163 std::size_t End = Body.size(), Pos = 0;
1164 for (; Pos != End; ++Pos) {
1165 // Check for a substitution or escape.
1166 if (Body[Pos] != '$' || Pos + 1 == End)
1169 char Next = Body[Pos + 1];
1170 if (Next == '$' || Next == 'n' || isdigit(Next))
1175 OS << Body.slice(0, Pos);
1177 // Check if we reached the end.
1181 switch (Body[Pos+1]) {
1187 // $n => number of arguments
1192 // $[0-9] => argument
1194 // Missing arguments are ignored.
1195 unsigned Index = Body[Pos+1] - '0';
1196 if (Index >= A.size())
1199 // Otherwise substitute with the token values, with spaces eliminated.
1200 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(),
1201 ie = A[Index].end(); it != ie; ++it)
1202 OS << it->getString();
1207 // Update the scan point.
1208 Body = Body.substr(Pos + 2);
1211 // We include the .endmacro in the buffer as our queue to exit the macro
1213 OS << ".endmacro\n";
1215 Instantiation = MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1218 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1220 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1221 // this, although we should protect against infinite loops.
1222 if (ActiveMacros.size() == 20)
1223 return TokError("macros cannot be nested more than 20 levels deep");
1225 // Parse the macro instantiation arguments.
1226 std::vector<std::vector<AsmToken> > MacroArguments;
1227 MacroArguments.push_back(std::vector<AsmToken>());
1228 unsigned ParenLevel = 0;
1230 if (Lexer.is(AsmToken::Eof))
1231 return TokError("unexpected token in macro instantiation");
1232 if (Lexer.is(AsmToken::EndOfStatement))
1235 // If we aren't inside parentheses and this is a comma, start a new token
1237 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1238 MacroArguments.push_back(std::vector<AsmToken>());
1240 // Adjust the current parentheses level.
1241 if (Lexer.is(AsmToken::LParen))
1243 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1246 // Append the token to the current argument list.
1247 MacroArguments.back().push_back(getTok());
1252 // Create the macro instantiation object and add to the current macro
1253 // instantiation stack.
1254 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1257 ActiveMacros.push_back(MI);
1259 // Jump to the macro instantiation and prime the lexer.
1260 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1261 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1267 void AsmParser::HandleMacroExit() {
1268 // Jump to the EndOfStatement we should return to, and consume it.
1269 JumpToLoc(ActiveMacros.back()->ExitLoc);
1272 // Pop the instantiation entry.
1273 delete ActiveMacros.back();
1274 ActiveMacros.pop_back();
1277 static void MarkUsed(const MCExpr *Value) {
1278 switch (Value->getKind()) {
1279 case MCExpr::Binary:
1280 MarkUsed(static_cast<const MCBinaryExpr*>(Value)->getLHS());
1281 MarkUsed(static_cast<const MCBinaryExpr*>(Value)->getRHS());
1283 case MCExpr::Target:
1284 case MCExpr::Constant:
1286 case MCExpr::SymbolRef: {
1287 static_cast<const MCSymbolRefExpr*>(Value)->getSymbol().setUsed(true);
1291 MarkUsed(static_cast<const MCUnaryExpr*>(Value)->getSubExpr());
1296 bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef) {
1297 // FIXME: Use better location, we should use proper tokens.
1298 SMLoc EqualLoc = Lexer.getLoc();
1300 const MCExpr *Value;
1301 if (ParseExpression(Value))
1306 if (Lexer.isNot(AsmToken::EndOfStatement))
1307 return TokError("unexpected token in assignment");
1309 // Eat the end of statement marker.
1312 // Validate that the LHS is allowed to be a variable (either it has not been
1313 // used as a symbol, or it is an absolute symbol).
1314 MCSymbol *Sym = getContext().LookupSymbol(Name);
1316 // Diagnose assignment to a label.
1318 // FIXME: Diagnostics. Note the location of the definition as a label.
1319 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1320 if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable())
1321 ; // Allow redefinitions of undefined symbols only used in directives.
1322 else if (!Sym->isUndefined() && (!Sym->isAbsolute() || !allow_redef))
1323 return Error(EqualLoc, "redefinition of '" + Name + "'");
1324 else if (!Sym->isVariable())
1325 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1326 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1327 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1330 // Don't count these checks as uses.
1331 Sym->setUsed(false);
1333 Sym = getContext().GetOrCreateSymbol(Name);
1335 // FIXME: Handle '.'.
1337 // Do the assignment.
1338 Out.EmitAssignment(Sym, Value);
1343 /// ParseIdentifier:
1346 bool AsmParser::ParseIdentifier(StringRef &Res) {
1347 // The assembler has relaxed rules for accepting identifiers, in particular we
1348 // allow things like '.globl $foo', which would normally be separate
1349 // tokens. At this level, we have already lexed so we cannot (currently)
1350 // handle this as a context dependent token, instead we detect adjacent tokens
1351 // and return the combined identifier.
1352 if (Lexer.is(AsmToken::Dollar)) {
1353 SMLoc DollarLoc = getLexer().getLoc();
1355 // Consume the dollar sign, and check for a following identifier.
1357 if (Lexer.isNot(AsmToken::Identifier))
1360 // We have a '$' followed by an identifier, make sure they are adjacent.
1361 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1364 // Construct the joined identifier and consume the token.
1365 Res = StringRef(DollarLoc.getPointer(),
1366 getTok().getIdentifier().size() + 1);
1371 if (Lexer.isNot(AsmToken::Identifier) &&
1372 Lexer.isNot(AsmToken::String))
1375 Res = getTok().getIdentifier();
1377 Lex(); // Consume the identifier token.
1382 /// ParseDirectiveSet:
1383 /// ::= .equ identifier ',' expression
1384 /// ::= .equiv identifier ',' expression
1385 /// ::= .set identifier ',' expression
1386 bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) {
1389 if (ParseIdentifier(Name))
1390 return TokError("expected identifier after '" + Twine(IDVal) + "'");
1392 if (getLexer().isNot(AsmToken::Comma))
1393 return TokError("unexpected token in '" + Twine(IDVal) + "'");
1396 return ParseAssignment(Name, allow_redef);
1399 bool AsmParser::ParseEscapedString(std::string &Data) {
1400 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1403 StringRef Str = getTok().getStringContents();
1404 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1405 if (Str[i] != '\\') {
1410 // Recognize escaped characters. Note that this escape semantics currently
1411 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1414 return TokError("unexpected backslash at end of string");
1416 // Recognize octal sequences.
1417 if ((unsigned) (Str[i] - '0') <= 7) {
1418 // Consume up to three octal characters.
1419 unsigned Value = Str[i] - '0';
1421 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1423 Value = Value * 8 + (Str[i] - '0');
1425 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1427 Value = Value * 8 + (Str[i] - '0');
1432 return TokError("invalid octal escape sequence (out of range)");
1434 Data += (unsigned char) Value;
1438 // Otherwise recognize individual escapes.
1441 // Just reject invalid escape sequences for now.
1442 return TokError("invalid escape sequence (unrecognized character)");
1444 case 'b': Data += '\b'; break;
1445 case 'f': Data += '\f'; break;
1446 case 'n': Data += '\n'; break;
1447 case 'r': Data += '\r'; break;
1448 case 't': Data += '\t'; break;
1449 case '"': Data += '"'; break;
1450 case '\\': Data += '\\'; break;
1457 /// ParseDirectiveAscii:
1458 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
1459 bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
1460 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1461 CheckForValidSection();
1464 if (getLexer().isNot(AsmToken::String))
1465 return TokError("expected string in '" + Twine(IDVal) + "' directive");
1468 if (ParseEscapedString(Data))
1471 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1473 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1477 if (getLexer().is(AsmToken::EndOfStatement))
1480 if (getLexer().isNot(AsmToken::Comma))
1481 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
1490 /// ParseDirectiveValue
1491 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1492 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1493 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1494 CheckForValidSection();
1497 const MCExpr *Value;
1498 if (ParseExpression(Value))
1501 // Special case constant expressions to match code generator.
1502 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value))
1503 getStreamer().EmitIntValue(MCE->getValue(), Size, DEFAULT_ADDRSPACE);
1505 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1507 if (getLexer().is(AsmToken::EndOfStatement))
1510 // FIXME: Improve diagnostic.
1511 if (getLexer().isNot(AsmToken::Comma))
1512 return TokError("unexpected token in directive");
1521 /// ParseDirectiveRealValue
1522 /// ::= (.single | .double) [ expression (, expression)* ]
1523 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1524 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1525 CheckForValidSection();
1528 // We don't truly support arithmetic on floating point expressions, so we
1529 // have to manually parse unary prefixes.
1531 if (getLexer().is(AsmToken::Minus)) {
1534 } else if (getLexer().is(AsmToken::Plus))
1537 if (getLexer().isNot(AsmToken::Integer) &&
1538 getLexer().isNot(AsmToken::Real))
1539 return TokError("unexpected token in directive");
1541 // Convert to an APFloat.
1542 APFloat Value(Semantics);
1543 if (Value.convertFromString(getTok().getString(),
1544 APFloat::rmNearestTiesToEven) ==
1545 APFloat::opInvalidOp)
1546 return TokError("invalid floating point literal");
1550 // Consume the numeric token.
1553 // Emit the value as an integer.
1554 APInt AsInt = Value.bitcastToAPInt();
1555 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1556 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1558 if (getLexer().is(AsmToken::EndOfStatement))
1561 if (getLexer().isNot(AsmToken::Comma))
1562 return TokError("unexpected token in directive");
1571 /// ParseDirectiveSpace
1572 /// ::= .space expression [ , expression ]
1573 bool AsmParser::ParseDirectiveSpace() {
1574 CheckForValidSection();
1577 if (ParseAbsoluteExpression(NumBytes))
1580 int64_t FillExpr = 0;
1581 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1582 if (getLexer().isNot(AsmToken::Comma))
1583 return TokError("unexpected token in '.space' directive");
1586 if (ParseAbsoluteExpression(FillExpr))
1589 if (getLexer().isNot(AsmToken::EndOfStatement))
1590 return TokError("unexpected token in '.space' directive");
1596 return TokError("invalid number of bytes in '.space' directive");
1598 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
1599 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
1604 /// ParseDirectiveZero
1605 /// ::= .zero expression
1606 bool AsmParser::ParseDirectiveZero() {
1607 CheckForValidSection();
1610 if (ParseAbsoluteExpression(NumBytes))
1614 if (getLexer().is(AsmToken::Comma)) {
1616 if (ParseAbsoluteExpression(Val))
1620 if (getLexer().isNot(AsmToken::EndOfStatement))
1621 return TokError("unexpected token in '.zero' directive");
1625 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
1630 /// ParseDirectiveFill
1631 /// ::= .fill expression , expression , expression
1632 bool AsmParser::ParseDirectiveFill() {
1633 CheckForValidSection();
1636 if (ParseAbsoluteExpression(NumValues))
1639 if (getLexer().isNot(AsmToken::Comma))
1640 return TokError("unexpected token in '.fill' directive");
1644 if (ParseAbsoluteExpression(FillSize))
1647 if (getLexer().isNot(AsmToken::Comma))
1648 return TokError("unexpected token in '.fill' directive");
1652 if (ParseAbsoluteExpression(FillExpr))
1655 if (getLexer().isNot(AsmToken::EndOfStatement))
1656 return TokError("unexpected token in '.fill' directive");
1660 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
1661 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
1663 for (uint64_t i = 0, e = NumValues; i != e; ++i)
1664 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
1669 /// ParseDirectiveOrg
1670 /// ::= .org expression [ , expression ]
1671 bool AsmParser::ParseDirectiveOrg() {
1672 CheckForValidSection();
1674 const MCExpr *Offset;
1675 if (ParseExpression(Offset))
1678 // Parse optional fill expression.
1679 int64_t FillExpr = 0;
1680 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1681 if (getLexer().isNot(AsmToken::Comma))
1682 return TokError("unexpected token in '.org' directive");
1685 if (ParseAbsoluteExpression(FillExpr))
1688 if (getLexer().isNot(AsmToken::EndOfStatement))
1689 return TokError("unexpected token in '.org' directive");
1694 // FIXME: Only limited forms of relocatable expressions are accepted here, it
1695 // has to be relative to the current section.
1696 getStreamer().EmitValueToOffset(Offset, FillExpr);
1701 /// ParseDirectiveAlign
1702 /// ::= {.align, ...} expression [ , expression [ , expression ]]
1703 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
1704 CheckForValidSection();
1706 SMLoc AlignmentLoc = getLexer().getLoc();
1708 if (ParseAbsoluteExpression(Alignment))
1712 bool HasFillExpr = false;
1713 int64_t FillExpr = 0;
1714 int64_t MaxBytesToFill = 0;
1715 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1716 if (getLexer().isNot(AsmToken::Comma))
1717 return TokError("unexpected token in directive");
1720 // The fill expression can be omitted while specifying a maximum number of
1721 // alignment bytes, e.g:
1723 if (getLexer().isNot(AsmToken::Comma)) {
1725 if (ParseAbsoluteExpression(FillExpr))
1729 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1730 if (getLexer().isNot(AsmToken::Comma))
1731 return TokError("unexpected token in directive");
1734 MaxBytesLoc = getLexer().getLoc();
1735 if (ParseAbsoluteExpression(MaxBytesToFill))
1738 if (getLexer().isNot(AsmToken::EndOfStatement))
1739 return TokError("unexpected token in directive");
1748 // Compute alignment in bytes.
1750 // FIXME: Diagnose overflow.
1751 if (Alignment >= 32) {
1752 Error(AlignmentLoc, "invalid alignment value");
1756 Alignment = 1ULL << Alignment;
1759 // Diagnose non-sensical max bytes to align.
1760 if (MaxBytesLoc.isValid()) {
1761 if (MaxBytesToFill < 1) {
1762 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
1763 "many bytes, ignoring maximum bytes expression");
1767 if (MaxBytesToFill >= Alignment) {
1768 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
1774 // Check whether we should use optimal code alignment for this .align
1776 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
1777 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
1778 ValueSize == 1 && UseCodeAlign) {
1779 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
1781 // FIXME: Target specific behavior about how the "extra" bytes are filled.
1782 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
1789 /// ParseDirectiveSymbolAttribute
1790 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
1791 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
1792 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1796 if (ParseIdentifier(Name))
1797 return TokError("expected identifier in directive");
1799 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1801 getStreamer().EmitSymbolAttribute(Sym, Attr);
1803 if (getLexer().is(AsmToken::EndOfStatement))
1806 if (getLexer().isNot(AsmToken::Comma))
1807 return TokError("unexpected token in directive");
1816 /// ParseDirectiveComm
1817 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
1818 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
1819 CheckForValidSection();
1821 SMLoc IDLoc = getLexer().getLoc();
1823 if (ParseIdentifier(Name))
1824 return TokError("expected identifier in directive");
1826 // Handle the identifier as the key symbol.
1827 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1829 if (getLexer().isNot(AsmToken::Comma))
1830 return TokError("unexpected token in directive");
1834 SMLoc SizeLoc = getLexer().getLoc();
1835 if (ParseAbsoluteExpression(Size))
1838 int64_t Pow2Alignment = 0;
1839 SMLoc Pow2AlignmentLoc;
1840 if (getLexer().is(AsmToken::Comma)) {
1842 Pow2AlignmentLoc = getLexer().getLoc();
1843 if (ParseAbsoluteExpression(Pow2Alignment))
1846 // If this target takes alignments in bytes (not log) validate and convert.
1847 if (Lexer.getMAI().getAlignmentIsInBytes()) {
1848 if (!isPowerOf2_64(Pow2Alignment))
1849 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
1850 Pow2Alignment = Log2_64(Pow2Alignment);
1854 if (getLexer().isNot(AsmToken::EndOfStatement))
1855 return TokError("unexpected token in '.comm' or '.lcomm' directive");
1859 // NOTE: a size of zero for a .comm should create a undefined symbol
1860 // but a size of .lcomm creates a bss symbol of size zero.
1862 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
1863 "be less than zero");
1865 // NOTE: The alignment in the directive is a power of 2 value, the assembler
1866 // may internally end up wanting an alignment in bytes.
1867 // FIXME: Diagnose overflow.
1868 if (Pow2Alignment < 0)
1869 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
1870 "alignment, can't be less than zero");
1872 if (!Sym->isUndefined())
1873 return Error(IDLoc, "invalid symbol redefinition");
1875 // '.lcomm' is equivalent to '.zerofill'.
1876 // Create the Symbol as a common or local common with Size and Pow2Alignment
1878 getStreamer().EmitZerofill(Ctx.getMachOSection(
1879 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
1880 0, SectionKind::getBSS()),
1881 Sym, Size, 1 << Pow2Alignment);
1885 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
1889 /// ParseDirectiveAbort
1890 /// ::= .abort [... message ...]
1891 bool AsmParser::ParseDirectiveAbort() {
1892 // FIXME: Use loc from directive.
1893 SMLoc Loc = getLexer().getLoc();
1895 StringRef Str = ParseStringToEndOfStatement();
1896 if (getLexer().isNot(AsmToken::EndOfStatement))
1897 return TokError("unexpected token in '.abort' directive");
1902 Error(Loc, ".abort detected. Assembly stopping.");
1904 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
1905 // FIXME: Actually abort assembly here.
1910 /// ParseDirectiveInclude
1911 /// ::= .include "filename"
1912 bool AsmParser::ParseDirectiveInclude() {
1913 if (getLexer().isNot(AsmToken::String))
1914 return TokError("expected string in '.include' directive");
1916 std::string Filename = getTok().getString();
1917 SMLoc IncludeLoc = getLexer().getLoc();
1920 if (getLexer().isNot(AsmToken::EndOfStatement))
1921 return TokError("unexpected token in '.include' directive");
1923 // Strip the quotes.
1924 Filename = Filename.substr(1, Filename.size()-2);
1926 // Attempt to switch the lexer to the included file before consuming the end
1927 // of statement to avoid losing it when we switch.
1928 if (EnterIncludeFile(Filename)) {
1929 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
1936 /// ParseDirectiveIf
1937 /// ::= .if expression
1938 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
1939 TheCondStack.push_back(TheCondState);
1940 TheCondState.TheCond = AsmCond::IfCond;
1941 if(TheCondState.Ignore) {
1942 EatToEndOfStatement();
1946 if (ParseAbsoluteExpression(ExprValue))
1949 if (getLexer().isNot(AsmToken::EndOfStatement))
1950 return TokError("unexpected token in '.if' directive");
1954 TheCondState.CondMet = ExprValue;
1955 TheCondState.Ignore = !TheCondState.CondMet;
1961 bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
1963 TheCondStack.push_back(TheCondState);
1964 TheCondState.TheCond = AsmCond::IfCond;
1966 if (TheCondState.Ignore) {
1967 EatToEndOfStatement();
1969 if (ParseIdentifier(Name))
1970 return TokError("expected identifier after '.ifdef'");
1974 MCSymbol *Sym = getContext().LookupSymbol(Name);
1977 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined());
1979 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined());
1980 TheCondState.Ignore = !TheCondState.CondMet;
1986 /// ParseDirectiveElseIf
1987 /// ::= .elseif expression
1988 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
1989 if (TheCondState.TheCond != AsmCond::IfCond &&
1990 TheCondState.TheCond != AsmCond::ElseIfCond)
1991 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
1993 TheCondState.TheCond = AsmCond::ElseIfCond;
1995 bool LastIgnoreState = false;
1996 if (!TheCondStack.empty())
1997 LastIgnoreState = TheCondStack.back().Ignore;
1998 if (LastIgnoreState || TheCondState.CondMet) {
1999 TheCondState.Ignore = true;
2000 EatToEndOfStatement();
2004 if (ParseAbsoluteExpression(ExprValue))
2007 if (getLexer().isNot(AsmToken::EndOfStatement))
2008 return TokError("unexpected token in '.elseif' directive");
2011 TheCondState.CondMet = ExprValue;
2012 TheCondState.Ignore = !TheCondState.CondMet;
2018 /// ParseDirectiveElse
2020 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
2021 if (getLexer().isNot(AsmToken::EndOfStatement))
2022 return TokError("unexpected token in '.else' directive");
2026 if (TheCondState.TheCond != AsmCond::IfCond &&
2027 TheCondState.TheCond != AsmCond::ElseIfCond)
2028 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
2030 TheCondState.TheCond = AsmCond::ElseCond;
2031 bool LastIgnoreState = false;
2032 if (!TheCondStack.empty())
2033 LastIgnoreState = TheCondStack.back().Ignore;
2034 if (LastIgnoreState || TheCondState.CondMet)
2035 TheCondState.Ignore = true;
2037 TheCondState.Ignore = false;
2042 /// ParseDirectiveEndIf
2044 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
2045 if (getLexer().isNot(AsmToken::EndOfStatement))
2046 return TokError("unexpected token in '.endif' directive");
2050 if ((TheCondState.TheCond == AsmCond::NoCond) ||
2051 TheCondStack.empty())
2052 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
2054 if (!TheCondStack.empty()) {
2055 TheCondState = TheCondStack.back();
2056 TheCondStack.pop_back();
2062 /// ParseDirectiveFile
2063 /// ::= .file [number] string
2064 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
2065 // FIXME: I'm not sure what this is.
2066 int64_t FileNumber = -1;
2067 SMLoc FileNumberLoc = getLexer().getLoc();
2068 if (getLexer().is(AsmToken::Integer)) {
2069 FileNumber = getTok().getIntVal();
2073 return TokError("file number less than one");
2076 if (getLexer().isNot(AsmToken::String))
2077 return TokError("unexpected token in '.file' directive");
2079 StringRef Filename = getTok().getString();
2080 Filename = Filename.substr(1, Filename.size()-2);
2083 if (getLexer().isNot(AsmToken::EndOfStatement))
2084 return TokError("unexpected token in '.file' directive");
2086 if (FileNumber == -1)
2087 getStreamer().EmitFileDirective(Filename);
2089 if (getStreamer().EmitDwarfFileDirective(FileNumber, Filename))
2090 Error(FileNumberLoc, "file number already allocated");
2096 /// ParseDirectiveLine
2097 /// ::= .line [number]
2098 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
2099 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2100 if (getLexer().isNot(AsmToken::Integer))
2101 return TokError("unexpected token in '.line' directive");
2103 int64_t LineNumber = getTok().getIntVal();
2107 // FIXME: Do something with the .line.
2110 if (getLexer().isNot(AsmToken::EndOfStatement))
2111 return TokError("unexpected token in '.line' directive");
2117 /// ParseDirectiveLoc
2118 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2119 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2120 /// The first number is a file number, must have been previously assigned with
2121 /// a .file directive, the second number is the line number and optionally the
2122 /// third number is a column position (zero if not specified). The remaining
2123 /// optional items are .loc sub-directives.
2124 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
2126 if (getLexer().isNot(AsmToken::Integer))
2127 return TokError("unexpected token in '.loc' directive");
2128 int64_t FileNumber = getTok().getIntVal();
2130 return TokError("file number less than one in '.loc' directive");
2131 if (!getContext().isValidDwarfFileNumber(FileNumber))
2132 return TokError("unassigned file number in '.loc' directive");
2135 int64_t LineNumber = 0;
2136 if (getLexer().is(AsmToken::Integer)) {
2137 LineNumber = getTok().getIntVal();
2139 return TokError("line number less than one in '.loc' directive");
2143 int64_t ColumnPos = 0;
2144 if (getLexer().is(AsmToken::Integer)) {
2145 ColumnPos = getTok().getIntVal();
2147 return TokError("column position less than zero in '.loc' directive");
2151 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2153 int64_t Discriminator = 0;
2154 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2156 if (getLexer().is(AsmToken::EndOfStatement))
2160 SMLoc Loc = getTok().getLoc();
2161 if (getParser().ParseIdentifier(Name))
2162 return TokError("unexpected token in '.loc' directive");
2164 if (Name == "basic_block")
2165 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2166 else if (Name == "prologue_end")
2167 Flags |= DWARF2_FLAG_PROLOGUE_END;
2168 else if (Name == "epilogue_begin")
2169 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2170 else if (Name == "is_stmt") {
2171 SMLoc Loc = getTok().getLoc();
2172 const MCExpr *Value;
2173 if (getParser().ParseExpression(Value))
2175 // The expression must be the constant 0 or 1.
2176 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2177 int Value = MCE->getValue();
2179 Flags &= ~DWARF2_FLAG_IS_STMT;
2180 else if (Value == 1)
2181 Flags |= DWARF2_FLAG_IS_STMT;
2183 return Error(Loc, "is_stmt value not 0 or 1");
2186 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2189 else if (Name == "isa") {
2190 SMLoc Loc = getTok().getLoc();
2191 const MCExpr *Value;
2192 if (getParser().ParseExpression(Value))
2194 // The expression must be a constant greater or equal to 0.
2195 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2196 int Value = MCE->getValue();
2198 return Error(Loc, "isa number less than zero");
2202 return Error(Loc, "isa number not a constant value");
2205 else if (Name == "discriminator") {
2206 if (getParser().ParseAbsoluteExpression(Discriminator))
2210 return Error(Loc, "unknown sub-directive in '.loc' directive");
2213 if (getLexer().is(AsmToken::EndOfStatement))
2218 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
2219 Isa, Discriminator);
2224 /// ParseDirectiveStabs
2225 /// ::= .stabs string, number, number, number
2226 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2227 SMLoc DirectiveLoc) {
2228 return TokError("unsupported directive '" + Directive + "'");
2231 /// ParseDirectiveCFIStartProc
2232 /// ::= .cfi_startproc
2233 bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef,
2234 SMLoc DirectiveLoc) {
2235 return getStreamer().EmitCFIStartProc();
2238 /// ParseDirectiveCFIEndProc
2239 /// ::= .cfi_endproc
2240 bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) {
2241 return getStreamer().EmitCFIEndProc();
2244 /// ParseRegisterOrRegisterNumber - parse register name or number.
2245 bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register,
2246 SMLoc DirectiveLoc) {
2249 if (getLexer().is(AsmToken::Percent)) {
2250 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc,
2253 Register = getContext().getTargetAsmInfo().getDwarfRegNum(RegNo, true);
2255 return getParser().ParseAbsoluteExpression(Register);
2260 /// ParseDirectiveCFIDefCfa
2261 /// ::= .cfi_def_cfa register, offset
2262 bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef,
2263 SMLoc DirectiveLoc) {
2264 int64_t Register = 0;
2265 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2268 if (getLexer().isNot(AsmToken::Comma))
2269 return TokError("unexpected token in directive");
2273 if (getParser().ParseAbsoluteExpression(Offset))
2276 return getStreamer().EmitCFIDefCfa(Register, Offset);
2279 /// ParseDirectiveCFIDefCfaOffset
2280 /// ::= .cfi_def_cfa_offset offset
2281 bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef,
2282 SMLoc DirectiveLoc) {
2284 if (getParser().ParseAbsoluteExpression(Offset))
2287 return getStreamer().EmitCFIDefCfaOffset(Offset);
2290 /// ParseDirectiveCFIDefCfaRegister
2291 /// ::= .cfi_def_cfa_register register
2292 bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef,
2293 SMLoc DirectiveLoc) {
2294 int64_t Register = 0;
2295 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2298 return getStreamer().EmitCFIDefCfaRegister(Register);
2301 /// ParseDirectiveCFIOffset
2302 /// ::= .cfi_off register, offset
2303 bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) {
2304 int64_t Register = 0;
2307 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2310 if (getLexer().isNot(AsmToken::Comma))
2311 return TokError("unexpected token in directive");
2314 if (getParser().ParseAbsoluteExpression(Offset))
2317 return getStreamer().EmitCFIOffset(Register, Offset);
2320 static bool isValidEncoding(int64_t Encoding) {
2321 if (Encoding & ~0xff)
2324 if (Encoding == dwarf::DW_EH_PE_omit)
2327 const unsigned Format = Encoding & 0xf;
2328 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
2329 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
2330 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
2331 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
2334 const unsigned Application = Encoding & 0x70;
2335 if (Application != dwarf::DW_EH_PE_absptr &&
2336 Application != dwarf::DW_EH_PE_pcrel)
2342 /// ParseDirectiveCFIPersonalityOrLsda
2343 /// ::= .cfi_personality encoding, [symbol_name]
2344 /// ::= .cfi_lsda encoding, [symbol_name]
2345 bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal,
2346 SMLoc DirectiveLoc) {
2347 int64_t Encoding = 0;
2348 if (getParser().ParseAbsoluteExpression(Encoding))
2350 if (Encoding == dwarf::DW_EH_PE_omit)
2353 if (!isValidEncoding(Encoding))
2354 return TokError("unsupported encoding.");
2356 if (getLexer().isNot(AsmToken::Comma))
2357 return TokError("unexpected token in directive");
2361 if (getParser().ParseIdentifier(Name))
2362 return TokError("expected identifier in directive");
2364 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2366 if (IDVal == ".cfi_personality")
2367 return getStreamer().EmitCFIPersonality(Sym, Encoding);
2369 assert(IDVal == ".cfi_lsda");
2370 return getStreamer().EmitCFILsda(Sym, Encoding);
2374 /// ParseDirectiveCFIRememberState
2375 /// ::= .cfi_remember_state
2376 bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal,
2377 SMLoc DirectiveLoc) {
2378 return getStreamer().EmitCFIRememberState();
2381 /// ParseDirectiveCFIRestoreState
2382 /// ::= .cfi_remember_state
2383 bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal,
2384 SMLoc DirectiveLoc) {
2385 return getStreamer().EmitCFIRestoreState();
2388 /// ParseDirectiveMacrosOnOff
2391 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
2392 SMLoc DirectiveLoc) {
2393 if (getLexer().isNot(AsmToken::EndOfStatement))
2394 return Error(getLexer().getLoc(),
2395 "unexpected token in '" + Directive + "' directive");
2397 getParser().MacrosEnabled = Directive == ".macros_on";
2402 /// ParseDirectiveMacro
2404 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
2405 SMLoc DirectiveLoc) {
2407 if (getParser().ParseIdentifier(Name))
2408 return TokError("expected identifier in directive");
2410 if (getLexer().isNot(AsmToken::EndOfStatement))
2411 return TokError("unexpected token in '.macro' directive");
2413 // Eat the end of statement.
2416 AsmToken EndToken, StartToken = getTok();
2418 // Lex the macro definition.
2420 // Check whether we have reached the end of the file.
2421 if (getLexer().is(AsmToken::Eof))
2422 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
2424 // Otherwise, check whether we have reach the .endmacro.
2425 if (getLexer().is(AsmToken::Identifier) &&
2426 (getTok().getIdentifier() == ".endm" ||
2427 getTok().getIdentifier() == ".endmacro")) {
2428 EndToken = getTok();
2430 if (getLexer().isNot(AsmToken::EndOfStatement))
2431 return TokError("unexpected token in '" + EndToken.getIdentifier() +
2436 // Otherwise, scan til the end of the statement.
2437 getParser().EatToEndOfStatement();
2440 if (getParser().MacroMap.lookup(Name)) {
2441 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
2444 const char *BodyStart = StartToken.getLoc().getPointer();
2445 const char *BodyEnd = EndToken.getLoc().getPointer();
2446 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
2447 getParser().MacroMap[Name] = new Macro(Name, Body);
2451 /// ParseDirectiveEndMacro
2454 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
2455 SMLoc DirectiveLoc) {
2456 if (getLexer().isNot(AsmToken::EndOfStatement))
2457 return TokError("unexpected token in '" + Directive + "' directive");
2459 // If we are inside a macro instantiation, terminate the current
2461 if (!getParser().ActiveMacros.empty()) {
2462 getParser().HandleMacroExit();
2466 // Otherwise, this .endmacro is a stray entry in the file; well formed
2467 // .endmacro directives are handled during the macro definition parsing.
2468 return TokError("unexpected '" + Directive + "' in file, "
2469 "no current macro definition");
2472 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
2473 getParser().CheckForValidSection();
2475 const MCExpr *Value;
2477 if (getParser().ParseExpression(Value))
2480 if (getLexer().isNot(AsmToken::EndOfStatement))
2481 return TokError("unexpected token in directive");
2483 if (DirName[1] == 's')
2484 getStreamer().EmitSLEB128Value(Value);
2486 getStreamer().EmitULEB128Value(Value);
2492 /// \brief Create an MCAsmParser instance.
2493 MCAsmParser *llvm::createMCAsmParser(const Target &T, SourceMgr &SM,
2494 MCContext &C, MCStreamer &Out,
2495 const MCAsmInfo &MAI) {
2496 return new AsmParser(T, SM, C, Out, MAI);